Using statistics Canada LifePaths microsimulation model to project the disability status of Canadian elderly

This study is part of a broader program of research aiming at producing a set of projections up to 2031 of the chronic homecare needs of Canadian elderly and the human resources required to assist them. The first objective of this paper is to assess the impact of an optimistic disability scenario on future numbers of elderly Canadian in poor health aged 75 years and over using Statistics Canada LifePaths microsimulation model. More specifically, a module to project future disability status of Canadian elderly was used to produce two scenarios: the Base LifePaths Disability scenario and the Healthy scenario in which it is assumed that additional years gained in life expectancy will be years without disability. We have calculated the proportion of the older population in poor health by linking disability status to marital status and surviving children.

The second objective of this study is to compare Canada with nine European countries (Belgium, Czech Republic, England and Wales, Finland, France, Germany, Italy, Netherlands, Portugal) who participate in the research program “Future Elderly LIving Conditions In Europe” (FELICIE). International comparisons are relevant because they allow the investigators to assess the specificities of Canadian aging. In this case, comparative data are available since two groups of researchers, one Canadian and one European, have been working in close collaboration in the field of the demography of aging. These research programs have now ended and the results have been widely disseminated within the national and international scientific community (see Carrière et al., 2007 and Gaymu et al., 2007). Both research programs produced microsimulated projections for the next 25 years that give comparable results on marital status, family network and support, and on living arrangements and needs for care. Here, the present research allows the health dimension to be comparable in order to use the same future health scenarios; in the first one the gain in life expectancy results in a proportional growth in the years in disability while in the second one the added years in life expectancy are years without disability.

In this paper, disability was defined using some attributes of the Health Utility Index (HUI). The HUI, based on the Comprehensive Health Status Measurements System (CHSMS), takes into account both the quantitative and qualitative aspects of health (Torrance et al., 1996). It provides information on the functional health of an individual using a series of attributes. For the purpose of this study, information on an individual’s mobility, dexterity, cognition, and pain and discomfort were used to develop four levels of disability (none, mild, moderate, and severe):

• No disability

• Mild disability:

  Mobility problems but do not need any help; Dexterity problems but do not need any help from someone else (may or may not use special equipments); Somewhat forgetful and slight difficulty in thinking; Moderate and/or severe pain prevents performing some or few tasks.

• Moderate disability:

  Requires wheel chair or mechanical support to walk; Dexterity problem and needs help to perform some tasks; Very forgetful and a lot of difficulty in thinking; Severe pain prevents performing most tasks.

• Severe disability:

  Cannot walk or needs help from others to walk; Dexterity problems and needs help for most or all tasks; Unable to remember or think.

The results in this paper are generated with a Canadian dynamic longitudinal microsimulation model called LifePaths. This model has been developed for several years by Statistics Canada in a programming language named Modgen. As described on the web site of Statistics Canada (Statistics Canada, 2004), this model is an overlapping cohort model that produces for each run a representative sample of the Canadian population. The oldest birth cohort represented in LifePaths was born in 1872. That year was chosen so that in the year 1971 the model would have a complete and representative set of all ages from newborns to the elderly. This means that starting in 1971 LifePaths produces cross-sectional annual tabulations that can be compared to historical data. The year 1971 is the first year for which high quality socio-demographic data were available in the form of a Census with contemporary design.

The life course of an individual, called a case, is simulated in LifePaths as a series of events that occur in continuous time (so they are not artificially restricted to annual intervals) using behavioural equations estimated from a large number of surveys and of historical micro-data sources. For example, the model use the National Population Health Survey (NPHS), School Leavers Survey (SLS), Labour Force Survey (LFS), National Graduate Survey (NGS), Family Historical Survey (FHS), General Social Survey (GSS), Census data, Historical Statistics of Canada and many others. All over the course of a simulation, LifePaths use the information coming form these data to keep updating its list of pending events to ensure that the next scheduled event is the one that currently has the shortest waiting time. This provides a straightforward way of dealing with competing events. Waiting times provide a unifying framework for representing decision-making. Probabilistic decisions can be implemented so that the choice among alternatives is determined by comparing two or more waiting times.

Moreover, a LifePaths simulation consists of a set of mutually independent cases. Each case contains exactly one dominant individual in the first generation. The spouse and children of the dominant individual are simulated as part of the case. They are created to satisfy the marriage and the fertility equations¹.

LifePaths takes into consideration a large number of variables representing events occurring in the life course of an individual. In the present research project, we have used a limited number of those variables:

• Age

• Sex

• Schooling level

• Region

• Marital status

• Age of spouse

• Place of birth

• Surviving children

• Disability (and institutionalization)

It is important to keep in mind that LifePaths is a dynamic model implying that some other variables not listed here can affect the results.

Another important concept in a model like LifePaths is the Monte Carlo variation. All cases simulated in this model are using waiting times that incorporate a stochastic component. It is one of the reasons why LifePaths can reproduce the diversity observed in real populations. On the other hand, this variability affects the reliability of aggregate tabular results in the same way results taken from a small sample would be affected if drawn from a large population. In order to avoid this kind of problem occurring when dealing with rare events, we decided to simulate as many as six million cases for both base and healthy scenarios.

One of the key variables of this project and of all studies concerning chronic homecare needs of the elderly is disability. The data taken into account to generate this variable in the LifePaths module that we used are coming from Canada’s National Population Health Survey (NPHS). The NPHS first cycle of data collection took place in 1994–1995 and will continue every second year thereafter for 20 years. The NPHS fulfilled both cross-sectional and longitudinal needs during its first three cycles, but with Cycle 4 (2000–2001) this survey became strictly a longitudinal one with two components: the household survey and the health care institutions survey (Rowe, 2006).

This study presents simulation results for the period 2001 to 2031. Year 2001 is our baseline because it was the last Canadian Census with most available results, and 2031 is corresponding to the year when the bulk of the baby-boomers will reach the age of 75, an age when physical and mental impairments rise rapidly.

The last part of this study is a comparison of the situation of Canada with FELICIE countries. One should be aware that the disability module in LifePaths is dynamic whereas it is static in the model of the European team. Static microsimulation models typically use static aging techniques, changing certain variables on the original micro-data file to produce a file with the demographic and economic characteristics expected in the future years. Person weights are modified to change the total population and the weighted characteristics of the population. Dynamic microsimulation models age each person in the micro-data file from one year to the next by probabilistically deciding whether or not that person will get married, get divorced, have a child, drop out of school, get a job, change jobs, become unemployed, retire, or die.

In the current version of LifePaths, there are no transition probabilities for some variables necessary to reach our research program objectives. Consequently, by using cross-sectional information, we have estimated the probabilities of classifying each factor/covariate pattern in the different categories of these variables. In this paper, however, it is not an issue because the projections do not involve these variables.

Indeed, the projections of Canadian chronic homecare needs (which is the main objective of our research programme) involve cross-sectional incorporation of parameters related to key variables. The left part of Figure 1 lists the characteristics needed for this cross-sectional part of the projections. The microsimulation model produces additional characteristics for each individual, but only the ones presented in Figure 1 are used. We first have to run the microsimulation to obtain the population by age, sex, schooling population, since married individuals are assumed to live with their spouse.

Figure 1

Download asset Open asset

To evaluate the health status of a population, demographers and epidemiologists have created some indicators based on life table’s parameters (Robine et al., 1991). Here, we use the Disability Free Life Expectancy (DFLE) The total Life Expectancy (LE), all health status combined, could be divided in two parts: DFLE and Life Expectancy level, region of residence, marital status, age of spouse, place of birth, and number of surviving children. We also need to differentiate who lives in an institution and who lives in a private household. Only the population living in private households is kept since by definition home care services are strictly aimed at this population. This first step provides a picture of the Canadian population for any given year up to 2031 according to the characteristics on the left of Figure 4. Probabilities of disability status need for assistance, living arrangement, receipt of assistance, and source of assistance are then applied to these populations in the sequence shown in Figure 1. As shown, there is an assumption that living arrangements are partly determined by the need for assistance. Thus, living with others is partially seen as a strategy to cope with a need for assistance related to a disability. This was done only for the non-married with Disability (LEWD). LEWD can be calculated by multiplying persons-years in the life table by disability rates by age using the Sullivan Method. Then, LE_x = DFLE_x + LEWD_x (x = age) (1)

Recent trends show that LE, as well as DFLE and LEWD, increases with time (Table 1 and Table 2). The healthy scenario keeps the LEWD, 11,3 years, at the same level throughout the projection period, meaning that all added years in life expectancy are disability free years. In this paper, all additional disability free years will be added to DFLE at age 45 as our target population is the elderly aged 75 + from year 2001 to 2031.

Even if there is no doubt that there are close links between mortality and morbidity, the first one is quite simple to measure while measuring the health of a person or a population is quite more complex. There is no unanimity about definitions of what we refer to when we say that a person is in good health or poor health.

However, many people agree that a measure of health should be closely linked to disability: the ability to independently perform activities of daily living (Cambois et al. 2005). To be in poor health then refers to a person that is in need of care and eventually formal services. Given the four levels of disability defined previously in Canadian surveys, it is usually admitted that those with no disability or with a mild disability are considered in good health, leaving in the poor health group those with a moderate or a severe disability and of those living in institutions.

Past trends in disability status are not very straightforward and most of the time it is impossible to assert if in the future we will face an extension or a compression of morbidity (Robine, 2005). In the present study, the parameters of our base LifePaths scenario are kept constant throughout the projected period, while in our healthy scenario the additional years gained in life expectancy are assumed to be healthy years which means a compression of morbidity.

In 2001, there are approximately two men for every three women in the population in poor health aged 75 years and over (Table 3). However regardless of the scenario an increase of men in poor health is expected to occur at a much faster pace than for women in the decades to come. One must keep in mind that the sex-related differential of life expectancy is expected to decrease. But the sharp increase in number of people in poor health from 2021 on is mostly due to the baby-boomers reaching the age of 75.

In order to have the healthy scenario become effective from 2001, the disability parameters have been modified from 2000, thus making small differences in the numbers of people in poor health.

The size of the family network of people in poor health is a key determinant of their probability to rely on formal home care services. Consequently, it is important to assess how many of those individuals can benefit of the presence of surviving children. In the present study, the use of microsimulations and the substantial numbers of simulated cases generate a much more complete portrait than what could have been obtained with a survey sample. In 2001, elderly men and women in poor health without surviving children show similar proportions; however the situation of men in the next decades is expected to evolve more rapidly than women. As mentioned above, most of the increase takes place between 2021 and 2031, period during which most baby-boomers will reach the age of 75 (Table 4).

Elderly people experiencing the highest risks of being dependent of formal sources of assistance are those with a reduced family network, that is people without neither a spouse nor surviving children. Even if people in such a situation represents slightly more than 15% of all the individuals aged 75 and over in poor heath, the numbers will increase sharply for men throughout the projection period, whereas the women situation will vary slowly and even improve before 2021 in the healthy scenario (Table 5).

Such an exercise for the Canadian population is very interesting per se, but an international comparison brings the specificities of the aging situation in Canada for the coming years into focus.

Regardless of the scenario, the growth of the Canadian total population aged 75 and over is expected to be substantially more important than in European countries for the period 2001–2031. More specifically, the Canadian elderly population will be multiplied by approximately 2.5 while in Europe it will increase by 1.5 (Figure 1 and Figure 2).

Figure 2

Download asset Open asset

One can observe that the numbers of elderly Canadian in poor health will significantly increase, especially from year 2021 on. In the base scenario, the growth in Canada will be twice as important during the period 2021–2031 compared to the two preceding decades.

Since the base scenario for Canada creates an expansion of morbidity without any modification of the disability parameters, this expansion is only due to the varying characteristics of the population at risk. The results for FELICIE countries are quite different because in the constant disability scenario it is assumed that the share of years lived in disability remains constant (Gaymu et al., 2007). Thus, we can clearly see that, in Figure 2, both curves for total elderly population and for elderly in poor health are absolutely the same since they are superposed.

For more detailed information on Statistics Canada LifePaths microsimulation model, see Statistics Canada (2004).

This analysis is based on Statistics Canada’s LifePaths microsimulation model. The assumptions and calculations underlying the simulation results were prepared by Yann Décarie and Jacques Légaré and the responsibility for the use and interpretation of these data is entirely that of the authors.

We would like to thank all people who have helped us in building this research project: Dr Geoff Rowe from Statistics Canada, Dr Janice Keefe from Mount Saint Vincent University, and all the members of our research team (Équipe de Recherche sur le Vieillissement des Populations – ERVIPOP). Financial support was provided by the Alzheimer Society of Canada (§ 0640) and the Canadian Institutes of Health Research ( MOP-81108).

1. Version of Record published: December 31, 2011 (version 1)

© 2011, Légaré and Décarie

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.